Editorial character sheet of the same fictional woman shown in three different scenes on…

Editorial character sheet of the same fictional woman shown in three different scenes on one canvas, consistency driven entirely by a locked text description: mid-30s, warm olive skin, a small crescent scar above the left eyebrow, sharp hazel eyes, chin-length copper hair with a blunt fringe, a thin silver nose ring, freckles across the nose. Panel 1: she sips coffee at a rain-streaked cafe window, soft grey daylight. Panel 2: same woman laughing in a sunlit wheat field, golden hour rim light. Panel 3: same woman under neon at night, teal and magenta reflections on her skin. Photoreal, 50mm, shallow focus, consistent facial geometry across all three panels, thin white gutters between panels." Recipe: Flux 2 dev + Turbo LoRA · 1536x1024 · seed 8476504476693375109 · 8 steps. Whole thing ran on my own 5090, ~70s, zero cloud spend. Where does GPT Image 2 pull ahead for you — bigger pose changes?